A Baroclinic Instability Test Case on an Anelastic ... · A Baroclinic Instability Test Case on an...

Post on 04-Jun-2020

3 views 0 download

transcript

A Baroclinic Instability Test Case on an Anelastic Dynamical Core Daniel Rothenberg1, Ross Heikes2

1Cornell University, Ithaca, NY, 2Colorado State University, Ft. Collins, CO

Although the familiar quasi-static system of equations filters sound waves, its assumption of hydrostatic balance tends to distort small-scale motions such s turbulence and convection. By neglecting the time-tendency portion of the continuity equation and considering small deviations from a hydrostatically balanced reference state, the ‘anelastic’ system of equations can be derived (Randall, 2010). This system still filters sound waves, but is non-hydrostatic. An anelastic model based on the CSU icosahedral grid was developed by Hiroaki Miura. The model solves the system of equations in vorticity-divergence form: Because this system is non-hydrostatic but still filters sound waves, it scales well and is suitable for studying both large-scale processes and smaller scale ones like turbulence.

Background Baroclinic Wave Simulations

This work has been supported by the National Science Foundation Science and Technology Center for Multi-Scale Modeling of Atmospheric Processes, managed by Colorado State University under cooperative agreement No. ATM-0425247.

Acknowledgments

Arakawa, A. and C.S. Konor, 2009: Unification of the Anelastic and

Quasi-Hydrostatic Systems of Equations. Mon. Wea. Rev., 137, 710-

726

Heikes, R., 2009: Icosahedral grid, Presentation. CMMAP July 2009

Team Meeting

Miura, H., 2009: An unified model (dynamics). Presentation,

CMMAP July 2009 Team Meeting

Jablonowski, C. and D.L. Williamson, 2006: A baroclinic instability

test case for atmospheric model dynamical cores. Q. J. R. Meteorol.

Soc., 132, 2943-2975

Randall, D., 2010: The Anelastic and Boussinesq Approximations.

Quick Studies, Colorado State University. Web.

References

A simple utility for viewing the model’s Fortran binary output

was developed using Python/MayaVi/TraitsUI. Built on the

Enthought Python Distribution, the tool is portable and has

been tested on Mac, Windows, and Linux.

Please visit http://bitbucket.org/counters/geodesic-plotter

for more information.

Visualization

- Implement further grid optimizations, such as the “spring”

method, to help reduce grid noise in the steady state

simulation.

- Investigate errors due to approximation of the Laplacian

operator by the multigrid solver in the model.

- Improve visualization toolkit and continue to develop

analysis tools for working with geodesic-gridded datasets.

Future Work

𝜕𝛿

𝜕𝑡= 𝛻ℎ ⋅ 𝜂𝛻ℎ𝜓 + 𝐉 𝜂, 𝜒 − 𝛻ℎ ⋅ 𝑤

𝜕𝐯𝐡

𝜕𝑧− 𝛻ℎ

2𝐾ℎ − 𝛻ℎ2 𝑐𝑝𝜃𝑝𝜋′ + 𝐹𝛿

𝜕𝜂

𝜕𝑡= −𝛻ℎ ⋅ 𝜂𝐯𝐡 − 𝛻h × 𝑤

𝜕𝐯𝐡

𝜕𝑧+ 𝐹𝜂

𝜕𝑤

𝜕𝑡= −𝐯𝐡 ⋅ 𝛻𝑤 + 𝑔

𝜃′

𝜃0−

𝜕

𝜕𝑧𝑐𝑝𝜃0𝜋′ + 𝐹𝑤

𝜕𝜃

𝜕𝑡= −𝐯𝐤 ⋅ 𝛻ℎ𝜃 − 𝑤

𝜕𝜃

𝜕𝑧+

𝑄

𝑐𝑝𝜋0+ 𝐹0

The steady-state and baroclinic wave test cases of Jablonowski and Williamson (2006) were performed on the anelastic dynamical core and on a hydrostatic dynamical core for reference. The anelastic model was run with the following grid and timestep parameters: Initially, the model is set-up with prescribed, balanced initial conditions. Left unperturbed, the simulation remains in a steady state. However, a perturbation can be superimposed in the zonal wind in the northern hemisphere to trigger the evolution a baroclinic wave over several days. For use in the anelastic model, a somewhat altered initial conditions were derived by Miura (2009), as seen below in comparison to Jablonowski’s:

Methodology

(r) # of Cells Grid Spacing (km) Timestep (s)

(6) 40,962 125.1 300

(7) 163,842 62.55 180

Colored = Anelastic , Black = Original; courtesy Miura (2009)

Sfc. Potential Temperature (K) Sfc. Relative Vorticity (10-5, 10-4 s-1) 250m Pressure (hPa)

- Performed on Z-grid anelastic dynamical core with a z-coordinate in the vertical; resolution was r6.

- By Day 6, a large error associated with the model’s grid appears in the output.

- Error accumulates over time and causes wave to break too early.

- Same numerical error occurs on both higher and lower resolution grids, in approximately the same manner.

- Error originates at the lowest model layer.

Steady State Simulations

Sfc. Relative Vorticity (10-4 s-1) Sfc. Potential Temperature (K)

- Error pattern corresponding to the geodesic grid’s major seams are observed in the model’s vertical velocity field.

- After 12 days, a pattern with wave-number 5 becomes visible in both the northern and southern hemispheres and grows over time.

- Symmetry in the observed pattern corresponds to the pentagonal cells in the model’s computational mesh.

- Using an optimized grid for the model tended to reduce the magnitude of the error and confine it to the original icosahedron’s edges.

For more information, contact the author at dar256@cornell.edu